Data Scientist Nanodegree¶

Convolutional Neural Networks¶

Project: Write an Algorithm for a Dog Identification App¶

This notebook walks you through one of the most popular Udacity projects across machine learning and artificial intellegence nanodegree programs. The goal is to classify images of dogs according to their breed.

If you are looking for a more guided capstone project related to deep learning and convolutional neural networks, this might be just it. Notice that even if you follow the notebook to creating your classifier, you must still create a blog post or deploy an application to fulfill the requirements of the capstone project.

Also notice, you may be able to use only parts of this notebook (for example certain coding portions or the data) without completing all parts and still meet all requirements of the capstone project.


In this notebook, some template code has already been provided for you, and you will need to implement additional functionality to successfully complete this project. You will not need to modify the included code beyond what is requested. Sections that begin with '(IMPLEMENTATION)' in the header indicate that the following block of code will require additional functionality which you must provide. Instructions will be provided for each section, and the specifics of the implementation are marked in the code block with a 'TODO' statement. Please be sure to read the instructions carefully!

In addition to implementing code, there will be questions that you must answer which relate to the project and your implementation. Each section where you will answer a question is preceded by a 'Question X' header. Carefully read each question and provide thorough answers in the following text boxes that begin with 'Answer:'. Your project submission will be evaluated based on your answers to each of the questions and the implementation you provide.

Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. Markdown cells can be edited by double-clicking the cell to enter edit mode.

The rubric contains optional "Stand Out Suggestions" for enhancing the project beyond the minimum requirements. If you decide to pursue the "Stand Out Suggestions", you should include the code in this IPython notebook.


Why We're Here¶

In this notebook, you will make the first steps towards developing an algorithm that could be used as part of a mobile or web app. At the end of this project, your code will accept any user-supplied image as input. If a dog is detected in the image, it will provide an estimate of the dog's breed. If a human is detected, it will provide an estimate of the dog breed that is most resembling. The image below displays potential sample output of your finished project (... but we expect that each student's algorithm will behave differently!).

Sample Dog Output

In this real-world setting, you will need to piece together a series of models to perform different tasks; for instance, the algorithm that detects humans in an image will be different from the CNN that infers dog breed. There are many points of possible failure, and no perfect algorithm exists. Your imperfect solution will nonetheless create a fun user experience!

The Road Ahead¶

We break the notebook into separate steps. Feel free to use the links below to navigate the notebook.

  • Step 0: Import Datasets
  • Step 1: Detect Humans
  • Step 2: Detect Dogs
  • Step 3: Create a CNN to Classify Dog Breeds (from Scratch)
  • Step 4: Use a CNN to Classify Dog Breeds (using Transfer Learning)
  • Step 5: Create a CNN to Classify Dog Breeds (using Transfer Learning)
  • Step 6: Write your Algorithm
  • Step 7: Test Your Algorithm

Step 0: Import Datasets¶

Import Dog Dataset¶

In the code cell below, we import a dataset of dog images. We populate a few variables through the use of the load_files function from the scikit-learn library:

  • train_files, valid_files, test_files - numpy arrays containing file paths to images
  • train_targets, valid_targets, test_targets - numpy arrays containing onehot-encoded classification labels
  • dog_names - list of string-valued dog breed names for translating labels
In [2]:
from sklearn.datasets import load_files       
from keras.utils import np_utils
import numpy as np
from glob import glob
import shutil

shutil.copytree

# define function to load train, test, and validation datasets
def load_dataset(path):
    data = load_files(path)
    dog_files = np.array(data['filenames'])
    dog_targets = np_utils.to_categorical(np.array(data['target']), 133)
    return dog_files, dog_targets

# load train, test, and validation datasets
train_files, train_targets = load_dataset('data/dog_images/train')
valid_files, valid_targets = load_dataset('data/dog_images/valid')
test_files, test_targets = load_dataset('data/dog_images/test')

# load list of dog names
dog_names = [item[20:-1] for item in sorted(glob("data/dog_images/train/*/"))]

# print statistics about the dataset
print('There are %d total dog categories.' % len(dog_names))
print('There are %s total dog images.\n' % len(np.hstack([train_files, valid_files, test_files])))
print('There are %d training dog images.' % len(train_files))
print('There are %d validation dog images.' % len(valid_files))
print('There are %d test dog images.'% len(test_files))
There are 133 total dog categories.
There are 8351 total dog images.

There are 6680 training dog images.
There are 835 validation dog images.
There are 836 test dog images.

Import Human Dataset¶

In the code cell below, we import a dataset of human images, where the file paths are stored in the numpy array human_files.

In [3]:
import random
random.seed(8675309)

# load filenames in shuffled human dataset
human_files = np.array(glob("data/lfw/*/*"))
random.shuffle(human_files)

# print statistics about the dataset
print('There are %d total human images.' % len(human_files))
There are 13233 total human images.
In [131]:
import pandas as pd

# exploratory analysis of the dog images
dog_dict = {}
for dog_name in dog_names:
    dog_name = dog_name.split('.')[-1]
    dog_dict[dog_name] = 0
for file in train_files:
    file_dog_breed = file.split('.')[-2].split('\\')[0]
    dog_dict[file_dog_breed] += 1

df_dog_breed = pd.DataFrame(list(dog_dict.values()), columns=['Count'], index=list(dog_dict.keys()))
df_dog_breed = df_dog_breed.sort_values('Count', ascending=False)
df_dog_breed.head()
Out[131]:
Count
Alaskan_malamute 77
Border_collie 74
Basset_hound 73
Dalmatian 71
Basenji 69
In [134]:
df_dog_breed.plot(kind='barh', figsize=(15, 50), legend=False, xlabel='Count', ylabel='Dog Breed')
Out[134]:
<AxesSubplot: xlabel='Count', ylabel='Dog Breed'>

Step 1: Detect Humans¶

We use OpenCV's implementation of Haar feature-based cascade classifiers to detect human faces in images. OpenCV provides many pre-trained face detectors, stored as XML files on github. We have downloaded one of these detectors and stored it in the haarcascades directory.

In the next code cell, we demonstrate how to use this detector to find human faces in a sample image.

In [4]:
import cv2                
import matplotlib.pyplot as plt                        
%matplotlib inline                               

# extract pre-trained face detector
face_cascade = cv2.CascadeClassifier('haarcascades/haarcascade_frontalface_alt.xml')

# load color (BGR) image
img = cv2.imread(human_files[3])
# convert BGR image to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

# find faces in image
faces = face_cascade.detectMultiScale(gray)

# print number of faces detected in the image
print('Number of faces detected:', len(faces))

# get bounding box for each detected face
for (x,y,w,h) in faces:
    # add bounding box to color image
    cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),2)
    
# convert BGR image to RGB for plotting
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

# display the image, along with bounding box
plt.imshow(cv_rgb)
plt.show()
Number of faces detected: 1

Before using any of the face detectors, it is standard procedure to convert the images to grayscale. The detectMultiScale function executes the classifier stored in face_cascade and takes the grayscale image as a parameter.

In the above code, faces is a numpy array of detected faces, where each row corresponds to a detected face. Each detected face is a 1D array with four entries that specifies the bounding box of the detected face. The first two entries in the array (extracted in the above code as x and y) specify the horizontal and vertical positions of the top left corner of the bounding box. The last two entries in the array (extracted here as w and h) specify the width and height of the box.

Write a Human Face Detector¶

We can use this procedure to write a function that returns True if a human face is detected in an image and False otherwise. This function, aptly named face_detector, takes a string-valued file path to an image as input and appears in the code block below.

In [5]:
# returns "True" if face is detected in image stored at img_path
def face_detector(img_path):
    img = cv2.imread(img_path)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    faces = face_cascade.detectMultiScale(gray)
    return len(faces) > 0

(IMPLEMENTATION) Assess the Human Face Detector¶

Question 1: Use the code cell below to test the performance of the face_detector function.

  • What percentage of the first 100 images in human_files have a detected human face?
  • What percentage of the first 100 images in dog_files have a detected human face?

Ideally, we would like 100% of human images with a detected face and 0% of dog images with a detected face. You will see that our algorithm falls short of this goal, but still gives acceptable performance. We extract the file paths for the first 100 images from each of the datasets and store them in the numpy arrays human_files_short and dog_files_short.

Answer: 99% of the first 100 images in human_files have a detected human face. 12% of the first 100 images in dog_files have a detected human face.

In [10]:
human_files_short = human_files[:100]
dog_files_short = train_files[:100]
# Do NOT modify the code above this line.

from PIL import Image
from tqdm import tqdm

## TODO: Test the performance of the face_detector algorithm 
## on the images in human_files_short and dog_files_short.

# test on human images
true_positive = 0
for file in human_files_short:
    if face_detector(file):
        true_positive += 1
    else: 
        # display falsely classified human image
        print("I don't see any human here?!")
        plt.show(plt.imshow(Image.open(file)))
        
print('Number of true positives (humans identified as humans): {} of 100'.format(true_positive))

# test on dog images
false_positive = 0
for file in dog_files_short:
    if face_detector(file):
        false_positive += 1
        # display falsely classified dog image
        print("This doesn't look like a dog to me?!")
        plt.show(plt.imshow(Image.open(file)))

        
print('Number of false positives (dogs identified as humans): {} of 100'.format(false_positive))
I don't see any human here?!
Number of true positives (humans identified as humans): 99 of 100
This doesn't look like a dog to me?!
This doesn't look like a dog to me?!
This doesn't look like a dog to me?!
This doesn't look like a dog to me?!
This doesn't look like a dog to me?!
This doesn't look like a dog to me?!
This doesn't look like a dog to me?!
This doesn't look like a dog to me?!
This doesn't look like a dog to me?!
This doesn't look like a dog to me?!
This doesn't look like a dog to me?!
This doesn't look like a dog to me?!
Number of false positives (dogs identified as humans): 12 of 100

Question 2: This algorithmic choice necessitates that we communicate to the user that we accept human images only when they provide a clear view of a face (otherwise, we risk having unneccessarily frustrated users!). In your opinion, is this a reasonable expectation to pose on the user? If not, can you think of a way to detect humans in images that does not necessitate an image with a clearly presented face?

Answer: No, this expectation is not reasonable, because not all pictures of humans have a clearly presented face. Another way to detect humans is to use a object detection model that was trained on identifying different parts of the human body and not only the face. One of these models is the "Faster RCN Inception V2 COCO Model", which was used in the following optional task.

We suggest the face detector from OpenCV as a potential way to detect human images in your algorithm, but you are free to explore other approaches, especially approaches that make use of deep learning :). Please use the code cell below to design and test your own face detection algorithm. If you decide to pursue this optional task, report performance on each of the datasets.

In [58]:
# Code adapted from https://gist.github.com/madhawav/1546a4b99c8313f06c0b2d7d7b4a09e2

import numpy as np
import tensorflow as tf
import cv2
import time


class DetectorAPI:
    def __init__(self, path_to_ckpt):
        self.path_to_ckpt = path_to_ckpt

        self.detection_graph = tf.Graph()
        with self.detection_graph.as_default():
            od_graph_def = tf.compat.v1.GraphDef()
            with tf.compat.v2.io.gfile.GFile(self.path_to_ckpt, 'rb') as fid:
                serialized_graph = fid.read()
                od_graph_def.ParseFromString(serialized_graph)
                tf.import_graph_def(od_graph_def, name='')

        self.default_graph = self.detection_graph.as_default()
        self.sess = tf.compat.v1.Session(graph=self.detection_graph)

        # Definite input and output Tensors for detection_graph
        self.image_tensor = self.detection_graph.get_tensor_by_name('image_tensor:0')
        # Each box represents a part of the image where a particular object was detected.
        self.detection_boxes = self.detection_graph.get_tensor_by_name('detection_boxes:0')
        # Each score represent how level of confidence for each of the objects.
        # Score is shown on the result image, together with the class label.
        self.detection_scores = self.detection_graph.get_tensor_by_name('detection_scores:0')
        self.detection_classes = self.detection_graph.get_tensor_by_name('detection_classes:0')
        self.num_detections = self.detection_graph.get_tensor_by_name('num_detections:0')

    def processFrame(self, image):
        # Expand dimensions since the trained_model expects images to have shape: [1, None, None, 3]
        image_np_expanded = np.expand_dims(image, axis=0)
        # Actual detection.
        (boxes, scores, classes, num) = self.sess.run(
            [self.detection_boxes, self.detection_scores, self.detection_classes, self.num_detections],
            feed_dict={self.image_tensor: image_np_expanded})

        im_height, im_width,_ = image.shape
        boxes_list = [None for i in range(boxes.shape[1])]
        for i in range(boxes.shape[1]):
            boxes_list[i] = (int(boxes[0,i,0] * im_height),
                        int(boxes[0,i,1]*im_width),
                        int(boxes[0,i,2] * im_height),
                        int(boxes[0,i,3]*im_width))

        return boxes_list, scores[0].tolist(), [int(x) for x in classes[0].tolist()], int(num[0])

    def close(self):
        self.sess.close()
        self.default_graph.close()
In [69]:
# threshold for the detection score
threshold = 0.95

# returns "True" if at least one human is detected in image stored at img_path
def human_detector(model, img_path):
    img = plt.imread(img_path)
    boxes, scores, classes, num = model.processFrame(img)
    for i in range(len(boxes)):
            # Class 1 represents human
            if classes[i] == 1 and scores[i] > threshold:
                return True
    return False
In [233]:
from IPython.utils import io
import PIL.Image

# Initiate pretrained Faster RCN Inception V2 COCO Model
model_path = 'faster_rcnn_inception_v2_coco_2018_01_28/frozen_inference_graph.pb'
model = DetectorAPI(path_to_ckpt=model_path)

# test on human images
true_positive = 0
for file in human_files_short:
    with io.capture_output() as captured:        
        faces = human_detector(model, file)
    if faces:
        true_positive += 1
    else: 
        # display falsely classified human image
        print("I don't see any human here?!")
        plt.show(plt.imshow(PIL.Image.open(file)))

print('Number of true positives (humans identified as humans): {} of 100'.format(true_positive))

# test on dog images
false_positive = 0
for file in dog_files_short:
    with io.capture_output() as captured:        
        faces = human_detector(model, file)
    if faces:
        false_positive += 1
        # display falsely classified dog image
        print("This doesn't look like a dog to me?!")
        plt.show(plt.imshow(PIL.Image.open(file)))
        
print('Number of false positives (dogs identified as humans): {} of 100'.format(false_positive))
Number of true positives (humans identified as humans): 100 of 100
This doesn't look like a dog to me?!
This doesn't look like a dog to me?!
This doesn't look like a dog to me?!
This doesn't look like a dog to me?!
This doesn't look like a dog to me?!
Number of false positives (dogs identified as humans): 5 of 100

Answer: The "Faster RCN Inception V2 COCO Model" shows a better performance for both datasets compared to the previously used OpenCV model. 100% of the first 100 images in human_files have a detected human. 5% of the first 100 images in dog_files have a detected human. In the latter case, the misclassified images indicate that most of them have both a human and one or multiple dogs showing. This special case of both object classes present in one image could be interesting to analyze for future work.


Step 2: Detect Dogs¶

In this section, we use a pre-trained ResNet-50 model to detect dogs in images. Our first line of code downloads the ResNet-50 model, along with weights that have been trained on ImageNet, a very large, very popular dataset used for image classification and other vision tasks. ImageNet contains over 10 million URLs, each linking to an image containing an object from one of 1000 categories. Given an image, this pre-trained ResNet-50 model returns a prediction (derived from the available categories in ImageNet) for the object that is contained in the image.

In [65]:
from keras.applications.resnet import ResNet50

# define ResNet50 model
ResNet50_model = ResNet50(weights='imagenet')
Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/resnet/resnet50_weights_tf_dim_ordering_tf_kernels.h5
102967424/102967424 [==============================] - 10s 0us/step

Pre-process the Data¶

When using TensorFlow as backend, Keras CNNs require a 4D array (which we'll also refer to as a 4D tensor) as input, with shape

$$ (\text{nb_samples}, \text{rows}, \text{columns}, \text{channels}), $$

where nb_samples corresponds to the total number of images (or samples), and rows, columns, and channels correspond to the number of rows, columns, and channels for each image, respectively.

The path_to_tensor function below takes a string-valued file path to a color image as input and returns a 4D tensor suitable for supplying to a Keras CNN. The function first loads the image and resizes it to a square image that is $224 \times 224$ pixels. Next, the image is converted to an array, which is then resized to a 4D tensor. In this case, since we are working with color images, each image has three channels. Likewise, since we are processing a single image (or sample), the returned tensor will always have shape

$$ (1, 224, 224, 3). $$

The paths_to_tensor function takes a numpy array of string-valued image paths as input and returns a 4D tensor with shape

$$ (\text{nb_samples}, 224, 224, 3). $$

Here, nb_samples is the number of samples, or number of images, in the supplied array of image paths. It is best to think of nb_samples as the number of 3D tensors (where each 3D tensor corresponds to a different image) in your dataset!

In [71]:
from keras.utils import image_utils
from tqdm import tqdm

def path_to_tensor(img_path):
    # loads RGB image as PIL.Image.Image type
    img = image_utils.load_img(img_path, target_size=(224, 224))
    # convert PIL.Image.Image type to 3D tensor with shape (224, 224, 3)
    x = image_utils.img_to_array(img)
    # convert 3D tensor to 4D tensor with shape (1, 224, 224, 3) and return 4D tensor
    return np.expand_dims(x, axis=0)

def paths_to_tensor(img_paths):
    list_of_tensors = [path_to_tensor(img_path) for img_path in tqdm(img_paths)]
    return np.vstack(list_of_tensors)

Making Predictions with ResNet-50¶

Getting the 4D tensor ready for ResNet-50, and for any other pre-trained model in Keras, requires some additional processing. First, the RGB image is converted to BGR by reordering the channels. All pre-trained models have the additional normalization step that the mean pixel (expressed in RGB as $[103.939, 116.779, 123.68]$ and calculated from all pixels in all images in ImageNet) must be subtracted from every pixel in each image. This is implemented in the imported function preprocess_input. If you're curious, you can check the code for preprocess_input here.

Now that we have a way to format our image for supplying to ResNet-50, we are now ready to use the model to extract the predictions. This is accomplished with the predict method, which returns an array whose $i$-th entry is the model's predicted probability that the image belongs to the $i$-th ImageNet category. This is implemented in the ResNet50_predict_labels function below.

By taking the argmax of the predicted probability vector, we obtain an integer corresponding to the model's predicted object class, which we can identify with an object category through the use of this dictionary.

In [67]:
from keras.applications.resnet import preprocess_input, decode_predictions

def ResNet50_predict_labels(img_path):
    # returns prediction vector for image located at img_path
    img = preprocess_input(path_to_tensor(img_path))
    return np.argmax(ResNet50_model.predict(img))

Write a Dog Detector¶

While looking at the dictionary, you will notice that the categories corresponding to dogs appear in an uninterrupted sequence and correspond to dictionary keys 151-268, inclusive, to include all categories from 'Chihuahua' to 'Mexican hairless'. Thus, in order to check to see if an image is predicted to contain a dog by the pre-trained ResNet-50 model, we need only check if the ResNet50_predict_labels function above returns a value between 151 and 268 (inclusive).

We use these ideas to complete the dog_detector function below, which returns True if a dog is detected in an image (and False if not).

In [68]:
### returns "True" if a dog is detected in the image stored at img_path
def dog_detector(img_path):
    prediction = ResNet50_predict_labels(img_path)
    return ((prediction <= 268) & (prediction >= 151)) 

(IMPLEMENTATION) Assess the Dog Detector¶

Question 3: Use the code cell below to test the performance of your dog_detector function.

  • What percentage of the images in human_files_short have a detected dog?
  • What percentage of the images in dog_files_short have a detected dog?

Answer: 0% of the first 100 images in human_files have a detected dog. 100% of the first 100 images in dog_files have a detected dog.

In [75]:
### TODO: Test the performance of the dog_detector function
### on the images in human_files_short and dog_files_short.
from IPython.utils import io

# test on human images
false_positive = 0
for file in human_files_short:
    with io.capture_output() as captured:        
        faces = dog_detector(file)
    if faces:
        false_positive += 1
        # display falsely classified human image
        print("I don't see any human here?!")
        plt.show(plt.imshow(Image.open(file)))

print('Number of false positives (humans identified as dogs): {} of 100'.format(false_positive))

# test on dog images
true_positive = 0
for file in dog_files_short:
    with io.capture_output() as captured:        
        faces = dog_detector(file)
    if faces:
        true_positive += 1
    else:
        # display falsely classified dog image
        print("This doesn't look like a dog to me?!")
        plt.show(plt.imshow(Image.open(file)))
        
print('Number of true positives (dogs identified as dogs): {} of 100'.format(true_positive))
Number of false positives (humans identified as dogs): 0 of 100
Number of true positives (dogs identified as dogs): 100 of 100

Step 3: Create a CNN to Classify Dog Breeds (from Scratch)¶

Now that we have functions for detecting humans and dogs in images, we need a way to predict breed from images. In this step, you will create a CNN that classifies dog breeds. You must create your CNN from scratch (so, you can't use transfer learning yet!), and you must attain a test accuracy of at least 1%. In Step 5 of this notebook, you will have the opportunity to use transfer learning to create a CNN that attains greatly improved accuracy.

Be careful with adding too many trainable layers! More parameters means longer training, which means you are more likely to need a GPU to accelerate the training process. Thankfully, Keras provides a handy estimate of the time that each epoch is likely to take; you can extrapolate this estimate to figure out how long it will take for your algorithm to train.

We mention that the task of assigning breed to dogs from images is considered exceptionally challenging. To see why, consider that even a human would have great difficulty in distinguishing between a Brittany and a Welsh Springer Spaniel.

Brittany Welsh Springer Spaniel

It is not difficult to find other dog breed pairs with minimal inter-class variation (for instance, Curly-Coated Retrievers and American Water Spaniels).

Curly-Coated Retriever American Water Spaniel

Likewise, recall that labradors come in yellow, chocolate, and black. Your vision-based algorithm will have to conquer this high intra-class variation to determine how to classify all of these different shades as the same breed.

Yellow Labrador Chocolate Labrador

We also mention that random chance presents an exceptionally low bar: setting aside the fact that the classes are slightly imabalanced, a random guess will provide a correct answer roughly 1 in 133 times, which corresponds to an accuracy of less than 1%.

Remember that the practice is far ahead of the theory in deep learning. Experiment with many different architectures, and trust your intuition. And, of course, have fun!

Pre-process the Data¶

We rescale the images by dividing every pixel in every image by 255.

In [76]:
from PIL import ImageFile                            
ImageFile.LOAD_TRUNCATED_IMAGES = True                 

# pre-process the data for Keras
train_tensors = paths_to_tensor(train_files).astype('float32')/255
valid_tensors = paths_to_tensor(valid_files).astype('float32')/255
test_tensors = paths_to_tensor(test_files).astype('float32')/255
100%|██████████| 6680/6680 [00:32<00:00, 206.79it/s]
100%|██████████| 835/835 [00:04<00:00, 181.72it/s]
100%|██████████| 836/836 [00:04<00:00, 207.94it/s]

(IMPLEMENTATION) Model Architecture¶

Create a CNN to classify dog breed. At the end of your code cell block, summarize the layers of your model by executing the line:

    model.summary()

We have imported some Python modules to get you started, but feel free to import as many modules as you need. If you end up getting stuck, here's a hint that specifies a model that trains relatively fast on CPU and attains >1% test accuracy in 5 epochs:

Sample CNN

Question 4: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. If you chose to use the hinted architecture above, describe why you think that CNN architecture should work well for the image classification task.

Answer: I decided to use the suggested architecture as a suitable starting point since it provides a reasonable number of layers, which in turn would mean a good trade-off between training duration and number of parameters. This could help to detect meaningful patterns in the otherwise difficult to distinguish characteristics of different dog breeds.

In addition to the convolution and pooling layers, I added a batch normalization pattern, wich help to handle internal covariate shift by normalizing the hidden representations gained during training, i.e., the output of the convolution and pooling layers. Moreover, I added a dropout layer between the global average pooling layer and the dense layer to randomly switch off 20% of the neurons of the model. This step helps to enhance the learning of our model.

The three convolution layers use the Relu Activation function, whereas a Softmax activation function was applied to the final dense layer.

In [159]:
from keras.layers import Conv2D, MaxPooling2D, GlobalAveragePooling2D, BatchNormalization
from keras.layers import Dropout, Flatten, Dense
from keras.models import Sequential

num_classes = len(dog_names)
model = Sequential()

### TODO: Define your architecture.
model.add(Conv2D(input_shape=(224, 224, 3), filters=16, kernel_size=2, activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(BatchNormalization())

model.add(Conv2D(filters = 32, kernel_size=2, activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(BatchNormalization())

model.add(Conv2D(filters = 64, kernel_size=2, activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(BatchNormalization())

model.add(GlobalAveragePooling2D())
model.add(Dropout(0.2))
model.add(Dense(num_classes, activation='softmax'))

model.summary()
Model: "sequential_19"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 conv2d_2009 (Conv2D)        (None, 223, 223, 16)      208       
                                                                 
 max_pooling2d_1029 (MaxPool  (None, 111, 111, 16)     0         
 ing2D)                                                          
                                                                 
 batch_normalization_12 (Bat  (None, 111, 111, 16)     64        
 chNormalization)                                                
                                                                 
 conv2d_2010 (Conv2D)        (None, 110, 110, 32)      2080      
                                                                 
 max_pooling2d_1030 (MaxPool  (None, 55, 55, 32)       0         
 ing2D)                                                          
                                                                 
 batch_normalization_13 (Bat  (None, 55, 55, 32)       128       
 chNormalization)                                                
                                                                 
 conv2d_2011 (Conv2D)        (None, 54, 54, 64)        8256      
                                                                 
 max_pooling2d_1031 (MaxPool  (None, 27, 27, 64)       0         
 ing2D)                                                          
                                                                 
 batch_normalization_14 (Bat  (None, 27, 27, 64)       256       
 chNormalization)                                                
                                                                 
 global_average_pooling2d_17  (None, 64)               0         
  (GlobalAveragePooling2D)                                       
                                                                 
 dropout_14 (Dropout)        (None, 64)                0         
                                                                 
 dense_1159 (Dense)          (None, 133)               8645      
                                                                 
=================================================================
Total params: 19,637
Trainable params: 19,413
Non-trainable params: 224
_________________________________________________________________

Compile the Model¶

In [160]:
model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])

(IMPLEMENTATION) Train the Model¶

Train your model in the code cell below. Use model checkpointing to save the model that attains the best validation loss.

You are welcome to augment the training data, but this is not a requirement.

In [161]:
from keras.callbacks import ModelCheckpoint  

### TODO: specify the number of epochs that you would like to use to train the model.

epochs = 5

### Do NOT modify the code below this line.

checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.from_scratch.hdf5', 
                               verbose=1, save_best_only=True)

model.fit(train_tensors, train_targets, 
          validation_data=(valid_tensors, valid_targets),
          epochs=epochs, batch_size=20, callbacks=[checkpointer], verbose=1)
Epoch 1/5
334/334 [==============================] - ETA: 0s - loss: 4.8636 - accuracy: 0.0163
Epoch 1: val_loss improved from inf to 4.91515, saving model to saved_models\weights.best.from_scratch.hdf5
334/334 [==============================] - 461s 1s/step - loss: 4.8636 - accuracy: 0.0163 - val_loss: 4.9152 - val_accuracy: 0.0072
Epoch 2/5
334/334 [==============================] - ETA: 0s - loss: 4.7409 - accuracy: 0.0272
Epoch 2: val_loss improved from 4.91515 to 4.78025, saving model to saved_models\weights.best.from_scratch.hdf5
334/334 [==============================] - 454s 1s/step - loss: 4.7409 - accuracy: 0.0272 - val_loss: 4.7803 - val_accuracy: 0.0228
Epoch 3/5
334/334 [==============================] - ETA: 0s - loss: 4.6804 - accuracy: 0.0346
Epoch 3: val_loss improved from 4.78025 to 4.69954, saving model to saved_models\weights.best.from_scratch.hdf5
334/334 [==============================] - 457s 1s/step - loss: 4.6804 - accuracy: 0.0346 - val_loss: 4.6995 - val_accuracy: 0.0263
Epoch 4/5
334/334 [==============================] - ETA: 0s - loss: 4.6238 - accuracy: 0.0400
Epoch 4: val_loss improved from 4.69954 to 4.68685, saving model to saved_models\weights.best.from_scratch.hdf5
334/334 [==============================] - 431s 1s/step - loss: 4.6238 - accuracy: 0.0400 - val_loss: 4.6869 - val_accuracy: 0.0407
Epoch 5/5
334/334 [==============================] - ETA: 0s - loss: 4.5832 - accuracy: 0.0451
Epoch 5: val_loss improved from 4.68685 to 4.61825, saving model to saved_models\weights.best.from_scratch.hdf5
334/334 [==============================] - 417s 1s/step - loss: 4.5832 - accuracy: 0.0451 - val_loss: 4.6182 - val_accuracy: 0.0479
Out[161]:
<keras.callbacks.History at 0x279e7d0e1f0>

Load the Model with the Best Validation Loss¶

In [162]:
model.load_weights('saved_models/weights.best.from_scratch.hdf5')

Test the Model¶

Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 1%.

In [163]:
from IPython.utils import io

# get index of predicted dog breed for each image in test set
with io.capture_output() as captured:        
    dog_breed_predictions = [np.argmax(model.predict(np.expand_dims(tensor, axis=0))) for tensor in test_tensors]

# report test accuracy
test_accuracy = 100*np.sum(np.array(dog_breed_predictions)==np.argmax(test_targets, axis=1))/len(dog_breed_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)
Test accuracy: 4.5455%

Step 4: Use a CNN to Classify Dog Breeds¶

To reduce training time without sacrificing accuracy, we show you how to train a CNN using transfer learning. In the following step, you will get a chance to use transfer learning to train your own CNN.

Obtain Bottleneck Features¶

In [164]:
bottleneck_features = np.load('bottleneck_features/DogVGG16Data.npz')
train_VGG16 = bottleneck_features['train']
valid_VGG16 = bottleneck_features['valid']
test_VGG16 = bottleneck_features['test']

Model Architecture¶

The model uses the the pre-trained VGG-16 model as a fixed feature extractor, where the last convolutional output of VGG-16 is fed as input to our model. We only add a global average pooling layer and a fully connected layer, where the latter contains one node for each dog category and is equipped with a softmax.

In [165]:
VGG16_model = Sequential()
VGG16_model.add(GlobalAveragePooling2D(input_shape=train_VGG16.shape[1:]))
VGG16_model.add(Dense(133, activation='softmax'))

VGG16_model.summary()
Model: "sequential_20"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 global_average_pooling2d_18  (None, 512)              0         
  (GlobalAveragePooling2D)                                       
                                                                 
 dense_1160 (Dense)          (None, 133)               68229     
                                                                 
=================================================================
Total params: 68,229
Trainable params: 68,229
Non-trainable params: 0
_________________________________________________________________

Compile the Model¶

In [166]:
VGG16_model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])

Train the Model¶

In [167]:
checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.VGG16.hdf5', 
                               verbose=1, save_best_only=True)

VGG16_model.fit(train_VGG16, train_targets, 
          validation_data=(valid_VGG16, valid_targets),
          epochs=20, batch_size=20, callbacks=[checkpointer], verbose=1)
Epoch 1/20
327/334 [============================>.] - ETA: 0s - loss: 8.1740 - accuracy: 0.2222
Epoch 1: val_loss improved from inf to 3.64412, saving model to saved_models\weights.best.VGG16.hdf5
334/334 [==============================] - 6s 17ms/step - loss: 8.0801 - accuracy: 0.2257 - val_loss: 3.6441 - val_accuracy: 0.4515
Epoch 2/20
329/334 [============================>.] - ETA: 0s - loss: 2.2443 - accuracy: 0.5938
Epoch 2: val_loss improved from 3.64412 to 2.41903, saving model to saved_models\weights.best.VGG16.hdf5
334/334 [==============================] - 3s 9ms/step - loss: 2.2381 - accuracy: 0.5949 - val_loss: 2.4190 - val_accuracy: 0.5892
Epoch 3/20
328/334 [============================>.] - ETA: 0s - loss: 1.2763 - accuracy: 0.7338
Epoch 3: val_loss improved from 2.41903 to 2.14024, saving model to saved_models\weights.best.VGG16.hdf5
334/334 [==============================] - 3s 8ms/step - loss: 1.2725 - accuracy: 0.7349 - val_loss: 2.1402 - val_accuracy: 0.6431
Epoch 4/20
333/334 [============================>.] - ETA: 0s - loss: 0.8347 - accuracy: 0.8095
Epoch 4: val_loss improved from 2.14024 to 2.06666, saving model to saved_models\weights.best.VGG16.hdf5
334/334 [==============================] - 3s 9ms/step - loss: 0.8358 - accuracy: 0.8093 - val_loss: 2.0667 - val_accuracy: 0.6455
Epoch 5/20
332/334 [============================>.] - ETA: 0s - loss: 0.5745 - accuracy: 0.8583
Epoch 5: val_loss improved from 2.06666 to 1.96039, saving model to saved_models\weights.best.VGG16.hdf5
334/334 [==============================] - 3s 8ms/step - loss: 0.5758 - accuracy: 0.8578 - val_loss: 1.9604 - val_accuracy: 0.6802
Epoch 6/20
328/334 [============================>.] - ETA: 0s - loss: 0.4034 - accuracy: 0.8951
Epoch 6: val_loss did not improve from 1.96039
334/334 [==============================] - 3s 8ms/step - loss: 0.4044 - accuracy: 0.8946 - val_loss: 2.1490 - val_accuracy: 0.6683
Epoch 7/20
331/334 [============================>.] - ETA: 0s - loss: 0.2990 - accuracy: 0.9180
Epoch 7: val_loss did not improve from 1.96039
334/334 [==============================] - 3s 8ms/step - loss: 0.3015 - accuracy: 0.9180 - val_loss: 1.9760 - val_accuracy: 0.7006
Epoch 8/20
328/334 [============================>.] - ETA: 0s - loss: 0.2136 - accuracy: 0.9416
Epoch 8: val_loss did not improve from 1.96039
334/334 [==============================] - 3s 8ms/step - loss: 0.2158 - accuracy: 0.9412 - val_loss: 1.9912 - val_accuracy: 0.7042
Epoch 9/20
329/334 [============================>.] - ETA: 0s - loss: 0.1751 - accuracy: 0.9500
Epoch 9: val_loss improved from 1.96039 to 1.87513, saving model to saved_models\weights.best.VGG16.hdf5
334/334 [==============================] - 3s 8ms/step - loss: 0.1758 - accuracy: 0.9497 - val_loss: 1.8751 - val_accuracy: 0.7186
Epoch 10/20
331/334 [============================>.] - ETA: 0s - loss: 0.1348 - accuracy: 0.9610
Epoch 10: val_loss did not improve from 1.87513
334/334 [==============================] - 3s 8ms/step - loss: 0.1345 - accuracy: 0.9609 - val_loss: 1.8958 - val_accuracy: 0.7138
Epoch 11/20
329/334 [============================>.] - ETA: 0s - loss: 0.1128 - accuracy: 0.9672
Epoch 11: val_loss did not improve from 1.87513
334/334 [==============================] - 3s 8ms/step - loss: 0.1125 - accuracy: 0.9672 - val_loss: 2.0558 - val_accuracy: 0.7198
Epoch 12/20
328/334 [============================>.] - ETA: 0s - loss: 0.0756 - accuracy: 0.9776
Epoch 12: val_loss did not improve from 1.87513
334/334 [==============================] - 3s 8ms/step - loss: 0.0764 - accuracy: 0.9772 - val_loss: 1.9874 - val_accuracy: 0.7269
Epoch 13/20
328/334 [============================>.] - ETA: 0s - loss: 0.0631 - accuracy: 0.9814
Epoch 13: val_loss did not improve from 1.87513
334/334 [==============================] - 3s 8ms/step - loss: 0.0638 - accuracy: 0.9811 - val_loss: 1.9903 - val_accuracy: 0.7138
Epoch 14/20
332/334 [============================>.] - ETA: 0s - loss: 0.0540 - accuracy: 0.9842
Epoch 14: val_loss did not improve from 1.87513
334/334 [==============================] - 3s 8ms/step - loss: 0.0537 - accuracy: 0.9843 - val_loss: 2.1298 - val_accuracy: 0.7138
Epoch 15/20
332/334 [============================>.] - ETA: 0s - loss: 0.0367 - accuracy: 0.9870
Epoch 15: val_loss did not improve from 1.87513
334/334 [==============================] - 3s 8ms/step - loss: 0.0365 - accuracy: 0.9871 - val_loss: 2.1526 - val_accuracy: 0.7305
Epoch 16/20
334/334 [==============================] - ETA: 0s - loss: 0.0356 - accuracy: 0.9889
Epoch 16: val_loss did not improve from 1.87513
334/334 [==============================] - 3s 8ms/step - loss: 0.0356 - accuracy: 0.9889 - val_loss: 2.0575 - val_accuracy: 0.7377
Epoch 17/20
332/334 [============================>.] - ETA: 0s - loss: 0.0299 - accuracy: 0.9911
Epoch 17: val_loss did not improve from 1.87513
334/334 [==============================] - 3s 8ms/step - loss: 0.0297 - accuracy: 0.9912 - val_loss: 2.0599 - val_accuracy: 0.7341
Epoch 18/20
333/334 [============================>.] - ETA: 0s - loss: 0.0273 - accuracy: 0.9931
Epoch 18: val_loss did not improve from 1.87513
334/334 [==============================] - 3s 8ms/step - loss: 0.0273 - accuracy: 0.9931 - val_loss: 2.1852 - val_accuracy: 0.7389
Epoch 19/20
328/334 [============================>.] - ETA: 0s - loss: 0.0241 - accuracy: 0.9933
Epoch 19: val_loss did not improve from 1.87513
334/334 [==============================] - 3s 8ms/step - loss: 0.0242 - accuracy: 0.9933 - val_loss: 2.1975 - val_accuracy: 0.7246
Epoch 20/20
331/334 [============================>.] - ETA: 0s - loss: 0.0234 - accuracy: 0.9938
Epoch 20: val_loss did not improve from 1.87513
334/334 [==============================] - 3s 8ms/step - loss: 0.0232 - accuracy: 0.9939 - val_loss: 2.1076 - val_accuracy: 0.7473
Out[167]:
<keras.callbacks.History at 0x279e9656d60>

Load the Model with the Best Validation Loss¶

In [168]:
VGG16_model.load_weights('saved_models/weights.best.VGG16.hdf5')

Test the Model¶

Now, we can use the CNN to test how well it identifies breed within our test dataset of dog images. We print the test accuracy below.

In [170]:
from IPython.utils import io

# get index of predicted dog breed for each image in test set
with io.capture_output() as captured: 
    VGG16_predictions = [np.argmax(VGG16_model.predict(np.expand_dims(feature, axis=0))) for feature in test_VGG16]

# report test accuracy
test_accuracy = 100*np.sum(np.array(VGG16_predictions)==np.argmax(test_targets, axis=1))/len(VGG16_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)
Test accuracy: 73.4450%

Predict Dog Breed with the Model¶

In [171]:
from extract_bottleneck_features import *

def VGG16_predict_breed(img_path):
    # extract bottleneck features
    bottleneck_feature = extract_VGG16(path_to_tensor(img_path))
    # obtain predicted vector
    predicted_vector = VGG16_model.predict(bottleneck_feature)
    # return dog breed that is predicted by the model
    return dog_names[np.argmax(predicted_vector)]

Step 5: Create a CNN to Classify Dog Breeds (using Transfer Learning)¶

You will now use transfer learning to create a CNN that can identify dog breed from images. Your CNN must attain at least 60% accuracy on the test set.

In Step 4, we used transfer learning to create a CNN using VGG-16 bottleneck features. In this section, you must use the bottleneck features from a different pre-trained model. To make things easier for you, we have pre-computed the features for all of the networks that are currently available in Keras:

  • VGG-19 bottleneck features
  • ResNet-50 bottleneck features
  • Inception bottleneck features
  • Xception bottleneck features

The files are encoded as such:

Dog{network}Data.npz

where {network}, in the above filename, can be one of VGG19, Resnet50, InceptionV3, or Xception. Pick one of the above architectures, download the corresponding bottleneck features, and store the downloaded file in the bottleneck_features/ folder in the repository.

(IMPLEMENTATION) Obtain Bottleneck Features¶

In the code block below, extract the bottleneck features corresponding to the train, test, and validation sets by running the following:

bottleneck_features = np.load('bottleneck_features/Dog{network}Data.npz')
train_{network} = bottleneck_features['train']
valid_{network} = bottleneck_features['valid']
test_{network} = bottleneck_features['test']
In [173]:
### TODO: Obtain bottleneck features from another pre-trained CNN.

bottleneck_features = np.load('bottleneck_features/DogInceptionV3Data.npz')

train_inceptionV3 = bottleneck_features['train']
valid_inceptionV3 = bottleneck_features['valid']
test_inceptionV3  = bottleneck_features['test']

(IMPLEMENTATION) Model Architecture¶

Create a CNN to classify dog breed. At the end of your code cell block, summarize the layers of your model by executing the line:

    <your model's name>.summary()

Question 5: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. Describe why you think the architecture is suitable for the current problem.

Answer: The following image shows the Inception V3 architecture [doi: 10.1109/SIBGRAPI.2018.00012]

Inception V3

In order to adapt this network to our dog breed classification use case, we need to add a global average pooling layer that uses the input shape of our training data and a final dense layer that uses the number of to be classified dog breeds as its output shape.

In [174]:
### TODO: Define your architecture.

inceptionV3_model = Sequential()

inceptionV3_model.add(GlobalAveragePooling2D(input_shape=train_inceptionV3.shape[1:]))
inceptionV3_model.add(Dense(133, activation='softmax'))

inceptionV3_model.summary()
Model: "sequential_21"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 global_average_pooling2d_19  (None, 2048)             0         
  (GlobalAveragePooling2D)                                       
                                                                 
 dense_1161 (Dense)          (None, 133)               272517    
                                                                 
=================================================================
Total params: 272,517
Trainable params: 272,517
Non-trainable params: 0
_________________________________________________________________

(IMPLEMENTATION) Compile the Model¶

In [176]:
### TODO: Compile the model.
inceptionV3_model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])

(IMPLEMENTATION) Train the Model¶

Train your model in the code cell below. Use model checkpointing to save the model that attains the best validation loss.

You are welcome to augment the training data, but this is not a requirement.

In [182]:
### TODO: Train the model.

checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.inceptionv3.hdf5', 
                               verbose=1, 
                               save_best_only=True
                              )

inceptionV3_model.fit(train_inceptionV3, train_targets,
                    validation_data=(valid_inceptionV3, valid_targets),
                    epochs=20,
                    batch_size=20,
                    callbacks=[checkpointer], 
                    verbose=1)
Epoch 1/20
334/334 [==============================] - ETA: 0s - loss: 0.5048 - accuracy: 0.8501
Epoch 1: val_loss improved from inf to 0.65294, saving model to saved_models\weights.best.inceptionv3.hdf5
334/334 [==============================] - 4s 11ms/step - loss: 0.5048 - accuracy: 0.8501 - val_loss: 0.6529 - val_accuracy: 0.8443
Epoch 2/20
333/334 [============================>.] - ETA: 0s - loss: 0.3804 - accuracy: 0.8857
Epoch 2: val_loss did not improve from 0.65294
334/334 [==============================] - 4s 11ms/step - loss: 0.3800 - accuracy: 0.8859 - val_loss: 0.6836 - val_accuracy: 0.8228
Epoch 3/20
330/334 [============================>.] - ETA: 0s - loss: 0.3038 - accuracy: 0.9058
Epoch 3: val_loss did not improve from 0.65294
334/334 [==============================] - 4s 11ms/step - loss: 0.3028 - accuracy: 0.9060 - val_loss: 0.6673 - val_accuracy: 0.8443
Epoch 4/20
333/334 [============================>.] - ETA: 0s - loss: 0.2441 - accuracy: 0.9234
Epoch 4: val_loss did not improve from 0.65294
334/334 [==============================] - 4s 11ms/step - loss: 0.2435 - accuracy: 0.9237 - val_loss: 0.6855 - val_accuracy: 0.8431
Epoch 5/20
334/334 [==============================] - ETA: 0s - loss: 0.2103 - accuracy: 0.9352
Epoch 5: val_loss did not improve from 0.65294
334/334 [==============================] - 4s 11ms/step - loss: 0.2103 - accuracy: 0.9352 - val_loss: 0.7262 - val_accuracy: 0.8419
Epoch 6/20
331/334 [============================>.] - ETA: 0s - loss: 0.1733 - accuracy: 0.9450
Epoch 6: val_loss did not improve from 0.65294
334/334 [==============================] - 4s 11ms/step - loss: 0.1726 - accuracy: 0.9449 - val_loss: 0.7402 - val_accuracy: 0.8467
Epoch 7/20
331/334 [============================>.] - ETA: 0s - loss: 0.1474 - accuracy: 0.9535
Epoch 7: val_loss did not improve from 0.65294
334/334 [==============================] - 4s 11ms/step - loss: 0.1491 - accuracy: 0.9530 - val_loss: 0.7771 - val_accuracy: 0.8503
Epoch 8/20
330/334 [============================>.] - ETA: 0s - loss: 0.1242 - accuracy: 0.9591
Epoch 8: val_loss did not improve from 0.65294
334/334 [==============================] - 4s 11ms/step - loss: 0.1250 - accuracy: 0.9588 - val_loss: 0.8855 - val_accuracy: 0.8395
Epoch 9/20
334/334 [==============================] - ETA: 0s - loss: 0.1043 - accuracy: 0.9662
Epoch 9: val_loss did not improve from 0.65294
334/334 [==============================] - 4s 11ms/step - loss: 0.1043 - accuracy: 0.9662 - val_loss: 0.7941 - val_accuracy: 0.8515
Epoch 10/20
329/334 [============================>.] - ETA: 0s - loss: 0.0899 - accuracy: 0.9717
Epoch 10: val_loss did not improve from 0.65294
334/334 [==============================] - 4s 11ms/step - loss: 0.0922 - accuracy: 0.9714 - val_loss: 0.8440 - val_accuracy: 0.8551
Epoch 11/20
333/334 [============================>.] - ETA: 0s - loss: 0.0792 - accuracy: 0.9733
Epoch 11: val_loss did not improve from 0.65294
334/334 [==============================] - 4s 12ms/step - loss: 0.0790 - accuracy: 0.9734 - val_loss: 0.7868 - val_accuracy: 0.8623
Epoch 12/20
330/334 [============================>.] - ETA: 0s - loss: 0.0680 - accuracy: 0.9786
Epoch 12: val_loss did not improve from 0.65294
334/334 [==============================] - 4s 12ms/step - loss: 0.0679 - accuracy: 0.9786 - val_loss: 0.8867 - val_accuracy: 0.8647
Epoch 13/20
332/334 [============================>.] - ETA: 0s - loss: 0.0639 - accuracy: 0.9797
Epoch 13: val_loss did not improve from 0.65294
334/334 [==============================] - 4s 12ms/step - loss: 0.0642 - accuracy: 0.9796 - val_loss: 0.9427 - val_accuracy: 0.8539
Epoch 14/20
330/334 [============================>.] - ETA: 0s - loss: 0.0549 - accuracy: 0.9839
Epoch 14: val_loss did not improve from 0.65294
334/334 [==============================] - 4s 12ms/step - loss: 0.0548 - accuracy: 0.9838 - val_loss: 0.9094 - val_accuracy: 0.8551
Epoch 15/20
331/334 [============================>.] - ETA: 0s - loss: 0.0464 - accuracy: 0.9866
Epoch 15: val_loss did not improve from 0.65294
334/334 [==============================] - 4s 13ms/step - loss: 0.0460 - accuracy: 0.9867 - val_loss: 0.9390 - val_accuracy: 0.8467
Epoch 16/20
334/334 [==============================] - ETA: 0s - loss: 0.0431 - accuracy: 0.9868
Epoch 16: val_loss did not improve from 0.65294
334/334 [==============================] - 4s 13ms/step - loss: 0.0431 - accuracy: 0.9868 - val_loss: 0.9936 - val_accuracy: 0.8419
Epoch 17/20
334/334 [==============================] - ETA: 0s - loss: 0.0372 - accuracy: 0.9888
Epoch 17: val_loss did not improve from 0.65294
334/334 [==============================] - 4s 12ms/step - loss: 0.0372 - accuracy: 0.9888 - val_loss: 0.9926 - val_accuracy: 0.8503
Epoch 18/20
330/334 [============================>.] - ETA: 0s - loss: 0.0359 - accuracy: 0.9898
Epoch 18: val_loss did not improve from 0.65294
334/334 [==============================] - 4s 13ms/step - loss: 0.0358 - accuracy: 0.9897 - val_loss: 0.9642 - val_accuracy: 0.8395
Epoch 19/20
334/334 [==============================] - ETA: 0s - loss: 0.0289 - accuracy: 0.9912
Epoch 19: val_loss did not improve from 0.65294
334/334 [==============================] - 4s 13ms/step - loss: 0.0289 - accuracy: 0.9912 - val_loss: 0.9913 - val_accuracy: 0.8611
Epoch 20/20
330/334 [============================>.] - ETA: 0s - loss: 0.0301 - accuracy: 0.9897
Epoch 20: val_loss did not improve from 0.65294
334/334 [==============================] - 4s 12ms/step - loss: 0.0297 - accuracy: 0.9898 - val_loss: 1.0155 - val_accuracy: 0.8515
Out[182]:
<keras.callbacks.History at 0x279e2af0220>

(IMPLEMENTATION) Load the Model with the Best Validation Loss¶

In [183]:
### TODO: Load the model weights with the best validation loss.
inceptionV3_model.load_weights('saved_models/weights.best.inceptionv3.hdf5')

(IMPLEMENTATION) Test the Model¶

Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 60%.

In [200]:
from IPython.utils import io

### TODO: Calculate classification accuracy on the test dataset.
with io.capture_output() as captured:
    inceptionV3_predictions = [np.argmax(inceptionV3_model.predict(np.expand_dims(feature, axis=0))) for feature in test_inceptionV3]

## report test accuracy
test_accuracy = 100*np.sum(np.array(inceptionV3_predictions)==np.argmax(test_targets, axis=1))/len(inceptionV3_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)
Test accuracy: 79.6651%

(IMPLEMENTATION) Predict Dog Breed with the Model¶

Write a function that takes an image path as input and returns the dog breed (Affenpinscher, Afghan_hound, etc) that is predicted by your model.

Similar to the analogous function in Step 5, your function should have three steps:

  1. Extract the bottleneck features corresponding to the chosen CNN model.
  2. Supply the bottleneck features as input to the model to return the predicted vector. Note that the argmax of this prediction vector gives the index of the predicted dog breed.
  3. Use the dog_names array defined in Step 0 of this notebook to return the corresponding breed.

The functions to extract the bottleneck features can be found in extract_bottleneck_features.py, and they have been imported in an earlier code cell. To obtain the bottleneck features corresponding to your chosen CNN architecture, you need to use the function

extract_{network}

where {network}, in the above filename, should be one of VGG19, Resnet50, InceptionV3, or Xception.

In [203]:
### TODO: Write a function that takes a path to an image as input
### and returns the dog breed that is predicted by the model.
from IPython.utils import io

def predict_dog_breed(img_path):
    # extract bottleneck features
    bottleneck_feature = extract_InceptionV3(path_to_tensor(img_path))
    
    # get predicted vector
    with io.capture_output() as captured:
        pred_vector = inceptionV3_model.predict(bottleneck_feature)
    
    # return predicted dog breed
    return dog_names[np.argmax(pred_vector)]

Step 6: Write your Algorithm¶

Write an algorithm that accepts a file path to an image and first determines whether the image contains a human, dog, or neither. Then,

  • if a dog is detected in the image, return the predicted breed.
  • if a human is detected in the image, return the resembling dog breed.
  • if neither is detected in the image, provide output that indicates an error.

You are welcome to write your own functions for detecting humans and dogs in images, but feel free to use the face_detector and dog_detector functions developed above. You are required to use your CNN from Step 5 to predict dog breed.

(IMPLEMENTATION) Write your Algorithm¶

In [236]:
### TODO: Write your algorithm.
### Feel free to use as many code cells as needed.

from IPython.display import Image, display
from IPython.utils import io

# Initiate pretrained Faster RCN Inception V2 COCO Model
model_path = 'faster_rcnn_inception_v2_coco_2018_01_28/frozen_inference_graph.pb'
model = DetectorAPI(path_to_ckpt=model_path)

def breed_classification(img_path):
    # plot the image to be classified
    display(Image(img_path, width=200, height=200))
    
    # get detections for both humans and dogs
    with io.capture_output() as captured:
        dog_detected = dog_detector(img_path)
        human_detected = human_detector(model, img_path)

    # predict if it is a dog
    if dog_detected == 1:
        print("Hello doggo, you seem to be a: ")
        with io.capture_output() as captured:
            predicted_breed = predict_dog_breed(img_path).partition('.')[-1]
        return predicted_breed
    
    # predict if it is a human
    elif human_detected == 1:
        print("Hello human being, you look a little bit like a: ")
        with io.capture_output() as captured:
            predicted_breed = predict_dog_breed(img_path).partition('.')[-1]
        return predicted_breed
    
    else:
        return print("No dogs or humans detected.")

Step 7: Test Your Algorithm¶

In this section, you will take your new algorithm for a spin! What kind of dog does the algorithm think that you look like? If you have a dog, does it predict your dog's breed accurately? If you have a cat, does it mistakenly think that your cat is a dog?

(IMPLEMENTATION) Test Your Algorithm on Sample Images!¶

Test your algorithm at least six images on your computer. Feel free to use any images you like. Use at least two human and two dog images.

Question 6: Is the output better than you expected :) ? Or worse :( ? Provide at least three possible points of improvement for your algorithm.

Answer: The output is better than I expected. Most of the sample images were classified correctly, both in terms of distinguishing between dogs and humans and also regarding identifying the correct dog breed. However, for our family dog Manfred (golden retriever) and one of the Labrador retrievers wrong predictions were made, where the erroneous prediction comes from a class with very low inter-class variations compared to the true breed (at least after a brief visual inspection).

Possible improvements:

  1. Use data augmentation to make our breed detection algorithm more robust to different sizes, locations and distortions.
  2. Use more epochs to train a better prediction model.
  3. Use more training images to learn characteristic features that might not be included in our limited dataset.
In [ ]:
## TODO: Execute your algorithm from Step 6 on
## at least 6 images on your computer.
## Feel free to use as many code cells as needed.
In [237]:
print(breed_classification("images\American_water_spaniel_00648.jpg"))
Hello doggo, you seem to be a: 
American_water_spaniel
In [238]:
print(breed_classification("images\Brittany_02625.jpg"))
Hello doggo, you seem to be a: 
Brittany
In [239]:
print(breed_classification("images\Curly-coated_retriever_03896.jpg"))
Hello doggo, you seem to be a: 
Curly-coated_retriever
In [240]:
print(breed_classification("images\golden_retriever_1.jpg"))
Hello doggo, you seem to be a: 
Golden_retriever

Now some images of our family dog ;-)

In [241]:
print(breed_classification("images\golden_retriever_2.jpg"))
Hello doggo, you seem to be a: 
Golden_retriever
In [242]:
print(breed_classification("images\golden_retriever_3.jpg"))
Hello doggo, you seem to be a: 
Kuvasz
In [243]:
print(breed_classification("images\golden_retriever_4.jpg"))
Hello doggo, you seem to be a: 
Golden_retriever
In [244]:
print(breed_classification("images\golden_retriever_5.jpg"))
Hello doggo, you seem to be a: 
Golden_retriever
In [245]:
print(breed_classification("images\golden_retriever_6.jpg"))
Hello doggo, you seem to be a: 
Kuvasz
In [246]:
print(breed_classification("images\Labrador_retriever_06449.jpg"))
Hello doggo, you seem to be a: 
Labrador_retriever
In [247]:
print(breed_classification("images\Labrador_retriever_06455.jpg"))
Hello doggo, you seem to be a: 
Chesapeake_bay_retriever
In [248]:
print(breed_classification("images\Labrador_retriever_06457.jpg"))
Hello doggo, you seem to be a: 
Labrador_retriever
In [254]:
print(breed_classification("images\sample_human_1.jpg"))
Hello human being, you look a little bit like a: 
Chesapeake_bay_retriever
In [255]:
print(breed_classification("images\sample_human_2.jpg"))
Hello human being, you look a little bit like a: 
Bull_terrier
In [256]:
print(breed_classification("images\sample_human_3.jpg"))
Hello human being, you look a little bit like a: 
Finnish_spitz
In [251]:
print(breed_classification("images\Welsh_springer_spaniel_08203.jpg"))
Hello doggo, you seem to be a: 
Welsh_springer_spaniel